Sales & pipeline · Workflow template

Cold-email every lead in your sheet

Point Carly at your lead sheet and she works down it — researching each lead on the web, writing a personalized opener from real facts, and sending it from your own mailbox.

Google Sheets

Gmail

Outlook

Build this in Carly Or copy the prompt ↓

The workflow at a glance

Trigger

Run manually

Tool

Read lead sheet

Foreach

Each lead

Tool

Web research

LLM

Write opener

Action

Send

How this workflow works, step by step

1 Trigger Run manually

The flow begins on run manually. Carly is listening for this event and kicks off the rest of the steps automatically.
2 Tool Read lead sheet

Carly pulls live data from Google Sheets — "Read lead sheet." Whatever's needed for the next step is fetched fresh, not cached.
3 Foreach Each lead

Carly loops over a list — each lead — and runs the next steps once per item. Works for leads, attendees, rows, files, anything iterable.
4 Tool Web research

Carly pulls live data from the connected tool — "Web research." Whatever's needed for the next step is fetched fresh, not cached.
5 LLM Write opener

An LLM step does the thinking: write opener. Carly picks the right model for the job and you can swap models per step if you need to.
6 Action Send

Finally Carly takes action in Gmail or Outlook: send. This is the thing that actually moves the world.

When to use this template

You want this work to happen consistently without anyone remembering to do it.
You're already using Google Sheets and Gmail, and the manual version of this is eating real minutes a week.
You want a visible canvas you can audit and edit — not a black-box agent doing things you can't see.

When not to use it

This task only happens once or twice a month — the setup cost won't pay back.
The decision in the middle of the flow needs human judgment every single time, not just on edge cases.
Your stack doesn't include any of the apps above and you don't want to wire a generic HTTP step.

Build it in 10 minutes — copy this prompt

Paste this in an email to and Carly will build the workflow for you.

Hey Carly — please build me this workflow.

Name: Cold-email every lead in your sheet

What it should do:
Point Carly at your lead sheet and she works down it — researching each lead on the web, writing a personalized opener from real facts, and sending it from your own mailbox.

Steps:
1. Trigger: Run manually
2. Tool: Read lead sheet
3. Foreach: Each lead
4. Tool: Web research
5. LLM: Write opener
6. Action: Send

Tools to wire up: Google Sheets, Gmail, Outlook

Set it live once you've got it working. Thanks!

Or open the dashboard →

Apps you'll connect

Carly authenticates these for you over OAuth when she sets up the workflow.

Google Sheets

Gmail

Outlook

Questions teams ask before building this

How long does this workflow take to set up?

Most teams have "Cold-email every lead in your sheet" running in under 10 minutes. Copy the prompt on this page into an email to carly@usecarly.com and Carly wires the workflow together for you, including authenticating Google Sheets.

Can I edit this template after Carly builds it?

Yes. Every workflow lives on the Carly canvas where you can rename steps, change conditions, swap LLM models, add filters, and re-route branches. Templates are starting points, not locked recipes.

What if I use a different tool than the ones in this template?

Carly connects to 200+ tools, plus a generic HTTP node for any API that doesn't have a native integration. Tell Carly your stack when you send the prompt and she'll pick the right connectors.

Does this run automatically once it's built?

Yes. The trigger at the start of the workflow fires automatically — on a schedule, on an inbound email, on a webhook, on a calendar event, whatever this template specifies. Carly handles the listening; you only step in when a step is gated on you.

Does this workflow cost anything to run?

Running the workflow itself — triggers, tool calls, branches, filters, waits — is free with no per-task fees. You only pay for the AI steps in this template (the LLM nodes that classify, write, or decide). Everything around them runs at no cost.